Support Website Contact Support Blog

How to improve the Point Cloud classification automatically

How to improve the Point Cloud classification in this case?

As you can see, most of the solar panels have been classified as “Ground” which made the DEM to me inaccurate.

What is the best way to solve this, without editing the Point Cloud manually?


I’d be super curious to know if this is possible as well. My company does a lot of solar array as-builts, and some of them want topography. It’s a bit time consuming manually selection every array and putting it into Disabled.

I’ve found that the Point Cloud Classification isn’t currently precise enough to be useful. Still, it is neat and I see potential with it.

Welcome to the community Michael!

Michael and Miguel,

It appears automatic point cloud classification struggles with your use case given the highly reflective surfaces of the solar panels, which seem to float above the ground with no structure along the sides for further reference. Maybe adding surfaces along the face would be quicker than reclassifying? For more info, you can refer to the article: How to draw a Surface in the rayCloud.

I hope this info provides some help.


I’ve tried just about everything. Manual editing is your best option. I create a separate classification called “TEMP.” I take strips of data along the rows where they are flat and try to get everything relatively quickly. It sucks, but it works. Global Mapper is another option.

Thanks for your reply @Jonathan_Dennis;

I am afraid this feature is not really useful for me due to time consuming. It would be the same as the manual classification of the point cloud. I can use those two features when the amount of panels were low, but it’s not realistic over huge surfaces with thousands of panels and arrays.

Any other feedback is welcome! TIA.